Cardiovascular diseases are the most common cause of death worldwide. One way to diagnose these diseases are long term ECGs, which can be interpreted with the help of data-driven methods. From a data protection perspective, such analyses should be performed on edge devices. However, given the constrained computing capabilities of these devices, the edge suitability of the deterministic and artificial intelligence models, which are used to analyze ECG signals, must be examined in advance.The aim of this paper is to provide an overview of the edge suitability of various preprocessing steps for ECG analysis, as well as presenting a benchmarking framework with standardized measurement methods for comparing the edge suitability. To this end, a formal and experimental research methodology is employed. The preprocessing models are evaluated by assessing their performance using two datasets. In addition, an evaluation is conducted regarding edge computing requirements, using five different computing systems.The model evaluation is showing that all tested algorithms are suitable for use in edge computing, as the model-specific memory and computing requirements for processing a 15-second ECG signal were consistently within the lowest threshold.

Automatic question generation (AQG) is increasingly used to scale assessments in higher education by reducing instructor workload and improving the efficiency of assessment design. However, existing systems, including those powered by Large Language Models (LLMs), often lack semantic grounding and curricular traceability. This limits their traceability and integration into learning analytics (LA) workflows, where precision is essential. This study introduces a modular pipeline that integrates curriculum knowledge graphs (KGs) with LLMs through structured prompting. The goal is to demonstrate how curriculum-grounded, traceable question generation can be operationalized through knowledge-graph LLM integration for use in engineering assessment. Specifically, we ask: How can knowledge graph-grounded generative AI enhance the traceability and contextual alignment of automated assessment in engineering education? We achieved this by first transferring the course content (specifically, Development of Embedded Systems using FPGAs and Data Science in Medicine) into a property knowledge graph, structured across multiple levels. This provides a semantic backbone on which the pipeline is built. For each target node, a Context-Background-Content (CBC) triplet was extracted using level-aware heuristics to frame the focal knowledge for item generation. A scaffolding, role-segmented prompt guided the LLM to generate assessment items. Each generated item was tagged with metadata, encoding its curricular lineage to ensure traceability and analytics readiness. Its modular design supports scalability across multiple courses and disciplines. Expert reviews indicate strong semantic alignment, clear questions, transparency in the generation process, and potential to enhance trust in AI-generated assessments. Anticipated extensions include integration with personalized learning paths and adaptive testing.

Portable electrocardiogram (ECG) systems are becoming increasingly popular, which is also driving demand for efficient real-time analyses that comply with data protection regulations. The field of edge computing offers novel approaches to signal processing at the point of data acquisition, reducing latency and enabling a higher level of data security. In this context, the FACE project aims to leverage the strengths of both edge computing and cloud computing to develop an innovative machine learning (ML) solution for analyzing long-term electrocardiograms. This contribution analyses various criteria for the application of machine learning models for ECG analysis in edge computing in terms of their applicability in the project. In addition, the determination of the necessary hardware will be based on the resulting criteria derived from the literature review.

While learning personalization offers great potential for learners, modern practices in higher education require a deeper consideration of domain models and learning contexts, to develop effective personalization algorithms. This paper introduces an innovative approach to higher education curriculum modelling that utilizes large language models (LLMs) for knowledge graph (KG) completion, with the goal of creating personalized learning-path recommendations. Our research focuses on modelling university subjects and linking their topics to corresponding domain models, enabling the integration of learning modules from different faculties and institutions in the student's learning path. Central to our approach is a collaborative process, where LLMs assist human experts in extracting high-quality, fine-grained topics from lecture materials. We develop a domain, curriculum, and user models for university modules and stakeholders. We implement this model to create the KG from two study modules: Embedded Systems and Development of Embedded Systems Using FPGA. The resulting KG structures the curriculum and links it to the domain models. We evaluate our approach through qualitative expert feedback and quantitative graph quality metrics. Domain experts validated the relevance and accuracy of the model, while the graph quality metrics measured the structural properties of our KG. Our results show that the LLM-assisted graph completion approach enhances the ability to connect related courses across disciplines to personalize the learning experience. Expert feedback also showed high acceptance of the proposed collaborative approach for concept extraction and classification.

In emergencies, treatment needs to be fast, accu-rate and patient-specific. For instance, in emergency scenarios, obstacles like treatment environments and medical difficulties can lead to bad outcomes for patients. Additionally, a drastic change of health vitals can force paramedics to shift to a different treatment in the ongoing treatment of the patient in order to save a patient's life. The KIRETT (engl.: 'Artificial intelligence in rescue operations ‘) demonstrator is developed to provide a rescue operator with a wrist-worn device, enabling treatment recommendation (with the help of knowledge graph) with situation detection models to improve the emergency treatment of a patient. This paper aims to provide a qualitative evaluation of the 2-days testing in the KIRETT project with the focus of knowledge graphs, knowledge fusion, and user-experience-design (UX-design).

Learning personalization has gained an ever-growing interest in recent years, due to its potential for offering tailored recommendations for the learner’s needs and goals. However, the decision to adopt the learning recommendation is not solely that of the AI-based recommender, but involves the learners themselves to a great extent, since they are investing the time and effort to follow the recommendation. In this research, we investigated the dynamics of the decision-making process between the human learner and the AI system. We introduce an explainable AI approach to support the human understanding of the intelligent recommendations, and thus their ability to collaboratively modify it, and make a final decision on adopting it. We implement a multimodal explanation approach, combining information from domain experts and large language models (LLMs) to provide a comprehensive overview of the recommendation reasoning. We test our system in a complex scenario within nursing training. Our findings point out the high acceptance rate of explained recommendations, and the role that our explanations played in supporting the agency of the learners over making an informed decision about the recommendation.

The increasing popularity of portable ECG systems and the growing demand for privacy-compliant, energy-efficient real-time analysis require new approaches to signal processing at the point of data acquisition. In this context, the edge domain is acquiring increasing importance, as it not only reduces latency times, but also enables an increased level of data security. The FACE project aims to develop an innovative machine learning solution for analysing long-term electrocardiograms that synergistically combines the strengths of edge and cloud computing. In this thesis, various pre-processing steps of ECG signals are analysed with regard to their applicability in the project. The selection of suitable methods in the edge area is based in particular on criteria such as energy efficiency, processing capability and real-time capability.

Objective Digital twins (DTs) emerged in the wake of Industry 4.0 and the creation of cyber-physical systems, motivated by the increased availability and variability of machine and sensor data. DTs are a concept to create a digital representation of a physical entity and imitate its behavior, while feeding real-world data to the digital counterpart, thus allowing enabling digital simulations related to the real-world entity. The availability of new data sources raises the potential for developing structured approaches for prediction and analysis. Similarly, in the field of medicine and digital healthcare, the collection of patient-focused data is rising. Medical DTs, a new concept of structured, exchangeable representations of knowledge, are increasingly used for capturing personal health, targeting specific illnesses, or addressing complex healthcare scenarios in hospitals. Methods This article surveys the current state-of-the-art in applying DTs in healthcare, and how these twins are generated to support smart, personalized medicine. These concepts are applied to a DT for a simulated health-monitoring scenario. Results The DT use case is implemented using AnyLogic multi-agent simulation, monitoring the patient’s personal health indicators and their development. Conclusion The results indicate both possibilities and challenges and provide important insights for future DT implementations in healthcare. They have the potential to optimize healthcare in various ways, such as providing patient-centered health-monitoring.

Recommendation systems are evolving rapidly, with increasing emphasis on deep and sophisticated model architectures to handle complex user-item interactions, resulting in larger models for better performance. However, the deployment of these models faces challenges due to high computing requirements and latency issues. To address this, a General Bidirectional framework (BiDiGen) is proposed, integrating user preferences and item similarities for personalized and context-aware recommendations. BiDiGen employs knowledge distillation to transfer pre-trained knowledge from large teacher models to smaller student models, striking a balance between efficiency and effectiveness. The framework utilizes a bidirectional approach where teachers and students mutually learn from each other, along with hypergraph construction to capture higher-order correlations efficiently. Additionally, a denoising generator is proposed for identifying relevant items, and a sampling scheme is suggested to determine knowledge transfer between teachers and students, ultimately training them with distillation and collaborative filtering losses.

The integration of vital signs in healthcare has witnessed a steady rise, promising health professionals to assist in their daily tasks to improve patient treatment. In life-threatening situations, like rescue operations, crucial decisions need to be made in the shortest possible amount of time to ensure that excellent treatment is provided during life-saving measurements. The integration of vital signs in the treatment holds the potential to improve time utilization for rescuers in such critical situations. They furthermore serve to support health professionals during the treatment with useful information and suggestions. To achieve such a goal, the KIRETT project serves to provide treatment recommendations and situation detection, combined on a wrist-worn wearable for rescue operations.This paper aims to present the significant role of vital signs in the improvement of decision-making during rescue operations and show their impact on health professionals and patients in need.

Healthcare and Medicine are under constant pressure to provide patient-driven medical expertise to ensure a fast and accurate treatment of the patient. In such scenarios, the diagnosis contains, the family history, long term medical data and a detailed consultation with the patient. In time-critical emergencies, such conversation and time-consuming elaboration are not possible. Rescue services need to provide fast, reliable treatments for the patient in need. With the help of modern technologies, like treatment recommendations, real-time vitals-monitoring, and situation detection through artificial intelligence (AI) a situation can be analyzed and supported in providing fast, accurate patient-data-driven medical treatments. In KIRETT, a wearable device is developed to support in such scenarios and presents a way to provide treatment recommendation in rescue services. The objective of this paper is to present the quantitative results of a two-day KIRETT evaluation (14 participants) to analyze the needs of rescue operators in healthcare.

Taxonomy mining plays an important role for organizing and structuring of data in Content Management Systems (CMS). In this paper, we propose a novel approach that leverages multidimensional knowledge representation (MKR) for taxonomy mining from text documents and enriching the extracted information via Large Language Model (LLM). The data originates from a Smart City project in Germany, which addresses housing, care and health for elderly people. The applied method involves the extraction of relevant keywords from text and the utilization of the MKR framework to analyze and represent the information. Results are provided for a context builder that utilizes GPT-4 to enrich the taxonomy. The enriched taxonomy is then used in a WordPress CMS for information search, structuring and tagging of the blog entries accordingly.

In the field of medicine and healthcare, the utilization of medical expertise, based on medical knowledge combined with patients’ health information is a life-critical challenge for patients and health professionals. The within-laying complexity and variety form the need for a united approach to gather, analyze, and utilize existing knowledge of medical treatments, and medical operations to provide the ability to present knowledge for the means of accurate patient-driven decision-making. One way to achieve this is the fusion of multiple knowledge sources in healthcare. It provides health professionals the opportunity to select from multiple contextual aligned knowledge sources which enables the support for critical decisions. This paper presents multiple conceptual models for knowledge fusion in the field of medicine, based on a knowledge graph structure. It will evaluate, how knowledge fusion can be enabled and presents how to integrate various knowledge sources into the knowledge graph for rescue operations.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

In the age of artificial intelligence (AI), providing learners with suitable and sufficient explanations of AI-based recommendation algorithm's output becomes essential to enable them to make an informed decision about it. However, the rapid development of AI approaches for educational recommendations and their explainability is not accompanied by an equal level of evidence-based experimentation to evaluate the learning effect of those explanations. To address this issue, we propose an experimental web-based tool for evaluating multimodal and large language model (LLM) based explainability approaches. Our tool provides a comprehensive set of modular, interactive, and customizable explainability elements, which researchers and educators can utilize to study the role of individual and hybrid explainability methods. We design a two-stage evaluation of the proposed tool, with learners and with educators. Our preliminary results from the first stage show high acceptance of the tool's components, user-friendliness, and an induced motivation to use the explanations for exploring more information about the recommendation.

In the era of personalized education, the provision of comprehensible explanations for learning recommendations is of a great value to enhance the learner's understanding and engagement with the recommended learning content. Large language models (LLMs) and generative AI in general have recently opened new doors for generating human-like explanations, for and along learning recommendations. However, their precision is still far away from acceptable in a sensitive field like education. To harness the abilities of LLMs, while still ensuring a high level of precision towards the intent of the learners, this paper proposes an approach to utilize knowledge graphs (KG) as a source of factual context, for LLM prompts, reducing the risk of model hallucinations, and safeguarding against wrong or imprecise information, while maintaining an application-intended learning context. We utilize the semantic relations in the knowledge graph to offer curated knowledge about learning recommendations. With domain-experts in the loop, we design the explanation as a textual template, which is filled and completed by the LLM. Domain experts were integrated in the prompt engineering phase as part of a study, to ensure that explanations include information that is relevant to the learner. We evaluate our approach quantitatively using Rouge-N and Rouge-L measures, as well as qualitatively with experts and learners. Our results show an enhanced recall and precision of the generated explanations compared to those generated solely by the GPT model, with a greatly reduced risk of generating imprecise information in the final learning explanation.

Student commitment towards a learning recommendation is not separable from their understanding of the reasons it was recommended to them; and their ability to modify it based on that understanding. Among explainability approaches, chatbots offer the potential to engage the student in a conversation, similar to a discussion with a peer or a mentor. The capabilities of chatbots, however, are still not sufficient to replace a human mentor, despite the advancements of generative AI (GenAI) and large language models (LLM). Therefore, we propose an approach to utilize chatbots as mediators of the conversation and sources of limited and controlled generation of explanations, to harvest the potential of LLMs while reducing their potential risks at the same time. The proposed LLM-based chatbot supports students in understanding learning-paths recommendations. We use a knowledge graph (KG) as a human-curated source of information, to regulate the LLM's output through defining its prompt's context. A group chat approach is developed to connect students with human mentors, either on demand or in cases that exceed the chatbot's pre-defined tasks. We evaluate the chatbot with a user study, to provide a proof-of-concept and highlight the potential requirements and limitations of utilizing chatbots in conversational explainability.

Learning path recommendations are essential to acquire skill sets needed to solve real-life challenges. However, the main source of information that recommendation systems (RS) use to generate personalized paths is user data rather than the challenge that the user faces. In this research, we propose a problem-based approach to generate learning-path recommendations using knowledge graphs (KG) to connect learning materials, and large language models (LLM) for natural-language understanding and topic extraction. We construct a KG of courses and digital badges through human- and machine-extracted relations. Our RS analyzes a challenge written by the learner, extracts learning goals needed to solve that challenge, and then implements a Markov decision process (MDP) to select the optimal learning path. The learning path is then explained utilizing the KG and the LLM. We evaluate our KG relations in comparison to expert-defined tags. We also evaluate the recommendations and their explanations with a use-case approach. Our preliminary results show the ability of the proposed system to connect courses from different domains, recommend corresponding paths to the challenge requirements, and assign relevant explanations accordingly.

During COVID-19 pandemic, and in the post-pandemic era, the dependency on digital learning tools and integrating them into regular teaching practices revealed several challenges and requirements from those tools. Such tools are needed to accommodate existing learning materials that teachers use, and deliver them to students who are learning remotely. In this paper, we demonstrate a mobile learning approach for supporting remote students and teachers. Our proposed approach utilizes a pedagogically informed design of exercise templates to construct learning materials as stand-alone exercises. A web application is designed to enable teachers of creating those exercises. On the learner’s side, a Chatbot-based mobile application is developed to ensure an easy and understandable interaction with the system for young students. Students could engage in a conversation with the Chatbot, solve exercises and get predefined teacher feedback on their solutions, all without a direct contact with the teacher who created those exercises. We implement our system and test it in three European contexts, in Spain, Greece and Germany. Our evaluation reveals high acceptance of the system’s ability to integrate existing materials to offer them to remote students; high student engagement with the learning Chatbot; and several lessons-learned from the user interaction with the system.